Blog - Konrad Kowalski (rootsher)
posts (112)
- The Layers of HTTP in the Browser: From fetch() to QUIC9/9
From fetch() to the First Byte
The whole lifecycle in one place: browser policies, cache, connection, wire format, headers and body. Only with that model does the Network panel answer…
- CPU in Kubernetes from the kernel up12/12
Anatomy of a CPU problem: from an understated request to p99 latency
One incident end to end: a 500m request, no limit, no throttling, and p99 climbing from 35 to 180 ms. The whole path of the mechanism.
- Part II: Standardizing AI in the SDLC Across the Organization4/10
RAG as shared organizational knowledge
When the standard is not in the repo, the agent reaches for RAG. For the agent RAG is usually a tool with a description, for the organization it is a…
- Part II: Standardizing AI in the SDLC Across the Organization3/10
Agent repo and workspace
A managed agent does not get a cloned repo, a configured git or working credentials. The agent repo describes how to carry out a task, the product repo…
- CPU in Kubernetes from the kernel up11/12
Noisy neighbor and CPU overcommit: when one Pod breaks another
Overcommit works until bursts become correlated. An understated request is not only about placement - it is a weaker position under contention.
- Part II: Standardizing AI in the SDLC Across the Organization2/10
Shared model access
Instead of separate keys, endpoints and quotas in every team, the agent points at a logical deployment and the organization's platform maps it to an approved…
- CPU in Kubernetes from the kernel up10/12
How to debug CPU in Kubernetes without guessing
Usage, throttling, contention and pressure are four different phenomena. A single CPU chart settles none of them - you need an order of operations.
- Part II: Standardizing AI in the SDLC Across the Organization1/10
From a Jira ticket to an agent run
The Implementer Agent starts from an explicit event, not from polling Jira. The trigger decides when to start the agent, the agent decides what to do once it…
- CPU in Kubernetes from the kernel up9/12
NUMA, SMT and cache: when hardware topology starts to matter
Two workloads can get the same CPU time and do different amounts of work. This is where the CPU abstraction ends and the physics of the machine begins.
- The Layers of HTTP in the Browser: From fetch() to QUIC8/9
HTTP/3 and QUIC: Same HTTP, Different Transport
QUIC removes transport-level head-of-line blocking and sets up a session more cheaply, but it does not fix the dependencies the frontend creates in its own…